The role of atherosclerosis in the acute coronary syndromes (ACS) that occur in patients with rheumatoid arthritis (RA) has not been quantified in detail. We undertook this study to determine the extent to which ACS are associated with carotid atherosclerosis in RA.
We prospectively ascertained ACS, defined as myocardial infarction, unstable angina, cardiac arrest, or death due to ischemic heart disease, in an RA cohort. We measured carotid atherosclerosis using high-resolution ultrasound. We used Cox proportional hazards models to estimate the association between ACS and atherosclerosis, adjusting for demographic features, cardiovascular (CV) risk factors, and RA manifestations.
We performed carotid ultrasound on 636 patients whom we followed up for 3,402 person-years. During this time, 84 patients experienced 121 new or recurrent ACS events, a rate of 3.5 ACS events per 100 patient-years (95% confidence interval [95% CI] 3.0–4.3). Among the 599 patients without a history of ACS, 66 incident ACS events occurred over 3,085 person-years, an incidence of 2.1 ACS events per 100 person-years (95% CI 1.7–2.7). The incidence of new ACS events per 100 patient-years was 1.1 (95% CI 0.6–1.7) among patients without plaque, 2.5 (95% CI 1.7–3.8) among patients with unilateral plaque, and 4.3 (95% CI 2.9–6.3) among patients with bilateral plaque. Covariates associated with incident ACS events independent of atherosclerosis included male sex, diabetes mellitus, and a cumulative glucocorticoid dose of ≥20 gm.
Atherosclerosis is strongly associated with ACS in RA. RA patients with carotid plaque, multiple CV risk factors (particularly diabetes mellitus or hypertension), many swollen joints, and a high cumulative dose of glucocorticoids, as well as RA patients who are men, are at high risk of ACS.
Rheumatoid arthritis (RA) is associated with an increased susceptibility to cardiovascular (CV) disease and related mortality (1–5). The physiopathology of the increased CV risk is not fully understood at this time. Known CV risk factors and systemic inflammation increase the propensity to CV, as they do in people without RA (6–9). These are thought to contribute to atherosclerosis (10), which in the general population, is a strong predictor of CV disease. However, there have been few efforts to quantify the role of atherosclerosis as a determinant of CV morbidity in RA (11). The objective of the current analysis was to estimate the contribution of atherosclerosis to acute coronary syndromes (ACS) in RA.
PATIENTS AND METHODS
From 1996 through 2001, we enrolled into a prospective study known as Outcome of Rheumatoid Arthritis Longitudinal Evaluation (ÓRALE) 779 consecutive patients with RA classified according to the 1987 revised criteria of the American College of Rheumatology (12). After a baseline evaluation at the recruitment site, we invited patients for annual followup assessments at a clinical research center. Between February 2000 and February 2003, we invited all patients for an additional visit to undergo a high-resolution B-mode carotid ultrasound evaluation.
CV risk factor assessment.
We recorded the CV risk factors at the baseline evaluation and at each annual followup visit. Medical records were reviewed thoroughly. We considered hypertension to be present if the diagnosis was recorded by a physician, if patients received antihypertensive medications, or if the systolic/diastolic blood pressure measured by us was ≥140/≥90 mm Hg. For obesity, we measured height and weight at each visit and calculated the body mass index (BMI; in kg/m2). Patients were considered obese if their BMI was ≥30. We considered diabetes mellitus and hypercholesterolemia to be present if either diagnosis was recorded by a physician in a medical record or if antidiabetic or lipid-lowering drugs were prescribed. In addition, we considered patients to have hypercholesterolemia if their fasting plasma cholesterol measured by us during a study visit was ever ≥200 mg/dl and to have diabetes if their fasting blood sugar was ≥126 mg/dl. We classified patients as current smokers if they continued to smoke at the time of the initial study visit and as former smokers if they had quit before that time.
At each visit, we assessed 28 joints for tenderness or pain on motion and swelling, and we assessed 48 joints for deformity (13). We also assessed patients for the presence or absence of subcutaneous nodules using a standardized protocol.
Assessment of disease activity.
In addition to tenderness and swelling in 28 joints, we measured the Westergren erythrocyte sedimentation rate at each visit. We used these 3 variables to calculate the Disease Activity Score in 28 joints (14).
Ascertainment of glucocorticoid use.
We assessed glucocorticoid exposure as described previously (15). Briefly, at each visit, patients were asked whether they were receiving glucocorticoids and, if so, to give the date they were first prescribed and the dose. We verified this information with their list of medications and by pharmacy and medical records. We estimated the cumulative oral glucocorticoid dosage by multiplying the current daily dose by the number of days since glucocorticoids were initiated. For alternate-day or other non-daily schedules, we averaged the dose to obtain the daily amount. We considered patients who received intraarticular, intramuscular, or oral glucocorticoids less often than monthly as not having received glucocorticoids. At each followup visit, we updated information on glucocorticoid use and dosage, and if the patient was still taking glucocorticoids, we calculated the cumulative dose since the previous visit. We did not consider dosage changes between study visits. The total cumulative glucocorticoid dose was calculated as the summed dose over each interval, expressed in prednisone equivalents. We stratified the cumulative glucocorticoid dose into various quantiles to determine if there was a threshold above which ACS increased significantly.
One technician performed a duplex scan of the carotid arteries in all patients, following a standardized vascular protocol developed for the Multi-Ethnic Study of Atherosclerosis (16). We used an ATL HDI 3000 high-resolution imaging machine with an L7-4 Transducer (Philips Medical Systems North America). The technician acquired 4 standardized B-mode images and a Doppler flow measurement from both sides of the neck. The first image was of the distal common carotid artery, and the 3 other images were centered on the site of maximum near and far wall thickening in the proximal internal carotid artery or carotid bulb. Results were recorded on Maxell Professional Super VHS tape (Hitachi-Maxell) and mailed to a central facility (Ultrasound Reading Center, New England Medical Center) for grading of the carotid artery intima-media thickness (IMT) and carotid plaque. At the Reading Center, the images were digitized at 30 frames per second, and arterial diameter fluctuations with the cardiac cycle were observed. Images were selected and read by a single, certified reader who was blinded with regard to the patient characteristics.
Carotid plaque was identified as a discrete projection of ≥50% from the adjacent wall into the vessel lumen. For the IMT, we measured in end diastole at each of the near and far walls of the right and left common carotid arteries and the near and far walls of the anterior oblique, lateral, and posterior oblique views of the right and left internal carotid arteries, for a total of 16 IMT measurements per person. Maximal IMTs of the common and internal carotid arteries were obtained by averaging the maximal measurement from the near and far walls at each projection, from the right and left sides. Then the composite maximal IMT was calculated by averaging the common and internal carotid maximal IMT values. The result is 1 IMT value per person, expressed in millimeters. Our study involved a single ultrasonographer and a single reader. Nevertheless, to assess the technique's reliability, our reader reread 50 images, and a different reader reread a separate set of 50 images. The intrareader intraclass correlation coefficient for IMT was 0.99, and the interreader coefficient was 0.94. For plaque, the intrareader κ was 1.0, while the interreader κ was 0.94.
Acute coronary syndromes.
ACS were defined as unstable angina, myocardial infarction (MI), cardiac arrest, or death with ischemic heart disease listed as the first or first underlying cause of death on the death certificate. At baseline and at each followup examination, a physician interviewed all patients about the occurrence of hospitalizations and comorbidity, with specific attention to ACS and other CV events, including coronary revascularization procedures. Patients provided a written release of medical records as part of the study protocol, and thus, complete records of all reported events were obtained and reviewed. Past events were ascertained initially at the baseline evaluation. Information on new ACS was updated at subsequent study visits. Area hospitals were contacted annually to determine if cohort members were hospitalized, in which case full medical records were obtained. All ACS events were ascertained with dates and confirmed by medical records. Deaths were identified from a variety of sources including next-of-kin, physicians, obituaries, and local health departments. Online mortality databases were searched monthly using ad hoc computer programs. All deaths were confirmed by death certificate. Events were systematically adjudicated using standardized criteria by a board-certified cardiologist (GLF) (17, 18), who was provided with results of invasive and noninvasive tests pertaining to the event, but who was blinded to other clinical characteristics, including the results of the carotid ultrasound. Events were considered incident if the patient had not experienced any previous ACS.
We compared patients' characteristics at the time of enrollment in the ÓRALE study, according to the presence or absence of a history of ACS, using t-tests or chi-square tests as indicated. We used a time-to-event approach to estimate the association between the carotid ultrasound findings (IMT or plaque) and the occurrence of ACS. All patients contributed observation time from the time of enrollment until their last study visit or the censoring date of December 31, 2004. We performed 2 analyses; in both of them, the observation time began on the date of enrollment in the ÓRALE study. In the first analysis, we studied the incidence of new ACS occurring during observation in patients who had never experienced a CV event prior to enrollment in the ÓRALE study. In the second analysis, we considered both new and recurrent events occurring during the observation period. In the latter of these analyses, we considered events that occurred before enrollment as past events, using an indicator variable. We did not consider nonfatal ACS that occurred after the date of the last visit or after the censoring date.
We plotted the survival function of patients grouped according to the presence of carotid plaque, using the Kaplan-Meier product limit technique and the log-rank chi-square test to examine significance (19). Our interest was in estimating the association between carotid IMT or plaque and ACS. We used Cox proportional hazards modeling to adjust for potential confounders, which were entered as covariates in multivariable models. The values for sex, race, smoking history, and the carotid IMT or plaque were fixed and did not vary over the observation period. All other variables were available as time-varying covariates. We considered covariates in 2 ways in separate models: as the nonvarying values obtained at the initial examination and as time-varying covariates (20). We also tested models that included quadratic terms for age and BMI. We used a stepwise selection approach to multivariable model building, beginning with a full model, removing variables if their association with ACS had significance at P ≥ 0.075, and reentering them if P ≤ 0.05. We used robust variance estimates in the recurrent-event models, adjusted for within-person clustering. We report 95% confidence intervals (95% CIs) for all estimates. All analyses were done using the Stata/SE 8.2 software package.
We have described the characteristics of the 779 ÓRALE cohort members in previous reports (10, 21). When we began the carotid ultrasound assessments in February 2000, 66 patients had died and 32 had moved away from San Antonio. This left 681 patients still eligible for ultrasound scanning. Of these, we could not establish contact with 12, 19 declined participation, and in 13, we did not obtain an ultrasound scan because our evaluation was done at their residences. We thus performed a high-resolution carotid ultrasound on 637 patients (93.5% of those eligible). In 1 patient, the image was not of sufficient quality for an IMT measurement, and that patient was omitted from analysis, leaving 636 patients with data for analyses. Table 1 shows the clinical characteristics of the patients, according to whether or not they had experienced incident or recurrent CV events.
Table 1. Baseline characteristics of 636 patients with RA according to occurrence of ACS*
By the censoring date of December 31, 2004, the 636 patients who had undergone a carotid ultrasound had accrued 3,402 person-years of observation. During this time, 84 of the patients experienced 121 ACS events, a rate of 3.5 per 100 person-years (95% CI 3.0–4.3). Of these, 71 were nonfatal MIs, 24 were episodes of unstable angina, 13 were deaths due to ischemic heart disease, 10 were fatal MIs, and 3 were fatal cardiac arrests.
The 599 patients who had no prior history of ACS accrued 3,085 person-years of observation, during which there were 66 incident ACS events, for an incidence of 2.1 ACS events per 100 person-years (95% CI 1.7–2.7). Of these, 42 were nonfatal MIs, 9 were episodes of unstable angina, 8 were deaths due to ischemic heart disease, 5 were fatal MIs, and 2 were fatal cardiac arrests.
Table 2 shows the number of events, observation period, and rate of ACS, according to the extent of carotid plaque. We classified patients as having unilateral plaque, bilateral plaque, or no plaque. The rate of events was higher in patients with more extensive plaque in a “dose-dependent” pattern. Compared with patients who were plaque-free, the rate of incident ACS among patients with unilateral plaque more than doubled, and among patients with bilateral plaque, the rate nearly quadrupled. The differences between patients with no plaque, unilateral plaque, and bilateral plaque were statistically significant. This pattern was accentuated when both new and recurrent events were considered together (Table 2).
Table 2. Rate of incident and all acute coronary syndromes in patients with rheumatoid arthritis, according to the extent of carotid plaque on ultrasound
Per 100 person-years. 95% CI = 95% confidence interval.
Figure 1 shows Kaplan-Meier curves illustrating the probability of survival free of incident ACS and all ACS during the period of observation. Both graphs suggest a progressively higher risk of ACS depending on the extent of carotid plaque. In both of these graphs, the survival function was plotted without adjusting for age or sex. After adjustment for these 2 variables, the plots changed little.
We performed multivariable Cox proportional hazards regression to examine the extent to which the plaque–ACS association shown in Table 2 was independent of confounders such as age, sex, and other covariates. In these analyses, we noted that the association between the carotid ultrasound variables and ACS varied little when baseline or time-varying covariates were included in the models. However, the covariates that were selected differed somewhat when baseline or time-varying values for the covariates were included in the models.
In Tables 3 and 4, we show the models that included baseline variables for the covariates. Table 3 shows baseline characteristics associated with the occurrence of incident ACS in patients among whom ACS had not previously occurred. For this analysis, there were a total of 66 incident ACS during observation. In the bivariate analysis, the 3 demographic variables were associated with incident ACS, as were diabetes mellitus and hypertension, the deformed joint count, cumulative glucocorticoid dose, and the carotid IMT. In stepwise multivariable models that did not include a carotid ultrasound variable, the variables independently associated with incident ACS were male sex, non-Hispanic white ethnicity, diabetes mellitus, hypertension, and cumulative glucocorticoid dose. When we added a carotid ultrasound variable to this model, the only variables that remained significantly associated with incident ACS were male sex, diabetes mellitus, and a cumulative glucocorticoid dose of at least 20 gm (Table 3). In this model, the hazard ratio associated with the IMT was 1.31 per SD difference (95% CI 1.02–1.70) (P ≤ 0.05). We reran this model using plaque instead of the IMT (results not shown). The hazard ratio for unilateral plaque was 2.02 (95% CI 1.07–3.84) and for bilateral plaque it was 2.96 (95% CI 1.55–5.62).
Table 3. Bivariate and stepwise multivariable analyses of baseline factors associated with incident acute coronary syndromes in 599 patients with RA who underwent carotid ultrasound*
Values are hazard ratios (95% confidence intervals). Sixty-six incident acute coronary syndromes occurred in 66 of 599 patients who had not previously experienced CV events. Variables were measured at the first visit. See Table 1 for definitions.
P ≤ 0.001.
P ≤ 0.05.
P ≤ 0.01.
Tested separately in models that did not include tender or swollen joint counts or ESR.
Table 4. Bivariate and stepwise multivariable analyses of baseline factors associated with incident or recurrent acute coronary syndromes among 636 patients with RA who underwent carotid ultrasound*
Without IMT or past CV events
With IMT and past CV events
Values are hazard ratios (95% confidence intervals). One hundred twenty-one acute coronary syndromes occurred after enrollment among 84 of the 636 patients. Variables were measured at the first visit. Confidence intervals were estimated using robust standard errors to adjust for within-patient correlation. See Table 1 for definitions.
P ≤ 0.001.
P ≤ 0.01.
P ≤ 0.05.
Tested separately in models that did not include tender or swollen joint counts or ESR.
In interpreting these findings on incident ACS, it is important to consider that with 66 events associated with a dichotomous exposure that was present in half of the patients, as was the case with plaque, the minimum detectable hazard ratio is ∼1.99 (22). Thus, these results do not rule out associations that are less strong between the other covariates that we studied and incident ACS. We did not include treatment with anti–tumor necrosis factor (anti-TNF) agents in the model because no CV events occurred among the patients receiving these agents at baseline (Table 1), which precludes Cox modeling. When we tested models using time-varying covariates, only male sex and hypertension were associated with incident ACS, both of which remained significant when carotid plaque or IMT was added to the model.
In Table 4, we examine the association between the same set of baseline predictors and ACS, considering both new and recurrent events. In this analysis, we did not exclude patients who had experienced CV events prior to entry into the ÓRALE study. Moreover, since patients who survived an ACS remained at risk of experiencing a second and subsequent events, we considered recurrent events as well, using robust standard errors to avoid Type I or false-positive errors. There were 121 such ACS events for analysis. In the bivariate analyses, variables associated with ACS events were demographic features, cumulative glucocorticoid dose of ≥20 gm, RA duration, all of the CV risk factors except for BMI, a past history of CV events, and the carotid IMT. In the multivariable analysis that did not include past CV events or the carotid IMT, the factors associated with ACS included age at first visit, male sex, diabetes mellitus, hypertension, the swollen joint count, and cumulative glucocorticoid dose. When we added past CV events and the carotid IMT to the model, all except for age remained independently associated with ACS (Table 4). We retested this same model after substituting carotid plaque for IMT, and the hazard ratio for unilateral plaque was 2.54 (95% CI 1.39–4.64). For bilateral plaque, the hazard ratio was 5.89 (95% CI 3.22–10.78).
We tested the above model for incident and recurrent ACS, using time-varying covariates, with results that differed somewhat from the baseline model. In the time-varying model, age, male sex, diabetes mellitus, hypertension, subcutaneous nodules, and the number of swollen joints were associated with ACS. However, upon adding past CV events and the carotid IMT to the model, only nodules and swollen joints remained associated with new and recurrent ACS, suggesting that their mechanism of association with ACS is independent of atherosclerosis. Figure 2 shows the incidence of ACS in the ÓRALE cohort and of MI in the Cardiovascular Health Study (23), according to IMT quintiles.
Our results suggest that atherosclerosis contributes significantly to ACS in RA. This may at first seem unsurprising, considering the effects of atherosclerosis in the general population. However, RA patients have a high rate of CV disease, a problem which remains incompletely understood. Until recently, even major monographs about RA did not mention atherosclerosis and attributed MI in RA patients to coronary arteritis (24). While it is still possible that this and other mechanisms could operate in RA, our findings suggest that atherosclerosis is a major factor in the CV complications of RA.
Despite only 66 incident ACS events for analysis, carotid atherosclerosis provided a strong signal that was readily detected in our study. The presence of plaque in both internal carotid arteries nearly quadrupled the incidence of new ACS compared with that in patients without carotid plaque. This was independent of all potential confounders available to us, including age, sex, and the remaining variables in Table 3. This supports the notion that atherosclerosis is in the causal pathway as a mediating variable between CV risk factors and ACS, rather than a confounder. The association between ACS and diabetes or hypertension was weakened when ultrasound findings were added to the model (Tables 3 and 4), supporting the hypothesis of a mediating role for atherosclerosis. That diabetes and hypertension nevertheless remained independently associated with ACS serves to refocus attention on their importance in determining the occurrence of CV events in RA (25).
The absence of the remaining variables from the predictive model should not be interpreted to mean that they are not associated with ACS, but rather that more outcome events (i.e., more statistical power) would be needed to detect their weaker signals. Alternatively, their absence from the predictive model could mean that their association with ACS is mediated through atherosclerosis, as measured by the carotid IMT. An example of the latter is shown in Table 4, in which age loses its association with ACS upon the addition of atherosclerosis markers to the model, suggesting that age is associated with ACS by the mechanism of atherosclerosis accrual over time.
In the analysis for multiple events, the effect of carotid atherosclerosis was even stronger than for incident events. Here, bilateral plaque raised the rate of events by a factor of almost 8. The multivariable analysis paralleled the incident event analysis in that atherosclerosis remained a strong predictor. This analysis also allowed us to estimate the effect of an alternative definition of atherosclerosis, based on the existence of CV events prior to enrollment in the ÓRALE cohort. This variable was also a strong predictor of the occurrence of ACS during observation.
Manifestations of RA and its treatment, including swollen joints and exposure to a high cumulative dose of prednisone, and, in time-varying models, nodules, were significantly associated with multiple ACS events. These variables were independent of atherosclerosis defined by the carotid IMT or plaque and were also independent of the history of CV events. This statistical independence suggests that they affect ACS occurrence through a pathway that may be biologically independent of atherosclerosis. Swollen joints and subcutaneous nodules are inflammatory and extraarticular features, respectively, of RA. Their association with ACS further propels inflammation and extraarticular disease into the spotlight of CV disease in RA.
Accumulating evidence suggests that systemic inflammation plays a role in atherogenesis and in CV morbidity and mortality (26–33). In addition to the high-grade inflammation seen in rheumatoid joints, blood from RA patients also displays a high concentration of mediators of inflammation, such as interleukin-1 and TNFα (34–38). These mediators up-regulate cell-mediated immunity, promoting inflammatory cell migration through the vascular endothelium, resulting in endothelial dysfunction (39, 40). This is a common finding in many chronic inflammatory disease states (41, 42) and has been demonstrated in patients with RA (43–45), including patients who are young and do not possess any classic risk factors for atherosclerosis (46). In a previous study by our group (11), RA manifestations were most strongly associated with atherosclerosis in younger patients, while in older patients, the traditional CV risk factors were more important. This suggests that systemic inflammation exerts its effects on the vasculature early in RA. In addition, increased extraarticular manifestations and a high number of involved joints, poor functional status, and rheumatoid nodules have consistently been found to predict CV-related mortality in RA (47–50). Our findings therefore suggest that subcutaneous nodules and increased swollen joint counts likely reflect systemic inflammation and may themselves be considered independent predictors of ACS in patients with RA.
It is of interest to consider whether the presence of atherosclerosis in RA patients is more predisposing to ACS than is the case in people without RA. Although a non-RA control group was not available to us to directly address this question, studies in the existing literature suggest that the predictive properties of the IMT are similar in both populations. In their meta-analysis of the ability of the IMT to predict CV events in the general population (51), Lorenz and colleagues report a pooled, adjusted, and standardized hazard ratio of 1.17 (95% CI 1.13–1.22) for MI associated with the IMT, well within the 95% CI of 1.02–1.70 that we found (Table 3). O'Leary and colleagues, using the same carotid ultrasound protocol that we used in the present study, reported the incidence of MI according to IMT quintiles in the Cardiovascular Health Study, which sampled community-dwelling adults from the general population (23). Figure 2 shows the data from that study and the present study side by side. Although we caution readers not to view this graph as a formal comparison because, among other differences, the outcome variables of the 2 studies were not identical, it suggests that the role of atherosclerosis in explaining CV events in RA is quantitatively similar to that observed in the general population.
In addition to the limited power of this study to detect weak associations posed by the small number of ACS events that occurred, our study is also subject to potential limitations due to event misclassification and lack of information about events that may have occurred after a patient's last visit. We were careful to avoid misclassification by confirming all self-reports with medical records, which were then adjudicated by a cardiologist. We did not consider nonfatal ACS after the last patient visit, nor did we consider event-free time after that date. Extending the observation time to include more events would likely increase statistical power, allowing us to detect weaker associations that did not reach significance in the present analysis.
Two conclusions are possible from the observations presented. First, atherosclerosis, as measured by carotid ultrasound, is an important predictor of ACS in RA. This suggests that interventions to prevent or reduce atherosclerosis have the potential to decrease the incidence of ACS in RA. Carotid ultrasound also has potential as a tool for stratifying CV risk in patients with RA and maybe in patients with other rheumatic diseases. The technology is ubiquitous, noninvasive, radiation-free, and reasonably priced. Moreover, the American College of Cardiology Foundation/American Heart Association have recommended ultrasound measurement of the carotid IMT as a reasonable tool for CV risk assessment in asymptomatic adults at intermediate risk (52).
Second, CV risk factors, RA manifestations, and their treatment are associated with ACS independently of the extent of atherosclerosis. This may implicate inflammatory mechanisms in the genesis of ACS in RA. For a better understanding of such mechanisms, further research is needed. Meanwhile, physicians who see patients with RA should be aware that those with diabetes mellitus, hypertension, high swollen joint counts, nodules, and high cumulative doses of glucocorticoids, as well as men with RA, are at high risk of acute coronary events.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. del Rincón had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Escalante, O'Leary, del Rincón.
Acquisition of data. Escalante, Battafarano, O'Leary, del Rincón.
Analysis and interpretation of data. Evans, Escalante, Battafarano, Freeman, del Rincón.